Conventionally, after an entire Multi-Chip Thin Film (MCM TF) module is completed, a full electrical test, an ATF test, is performed to confirm the integrity of the completed wiring. If any defect is detected at this stage, an after thin film (ATF) repair using top-surface repair lines is performed to correct the defective nets.
FIG. 1 shows a plan view of a typical MCM 100. In FIG. 1, chips 102, 104, 106, 108, 110, 112, and 114 are mounted to the top surface metallurgy (TSM) of MCM 100 using a Controlled-Collapsed-Chip-Connection (C4) configuration (not shown in this Figure). In FIG. 1 seven chip locations are shown. MCMs are not limited to this configuration, however, and may be any number of chips depending on the requirements of the application. Before mounting the chips 102 through 114, MCM 100 is tested to ensure that no open circuits or short circuits exist in MCM 100. If open circuits or short circuits are found, the MCM must be repaired.
The conventional ATF repair strategy discards the entire original net wiring and reconstructs new net wiring using the top surface repair lines, modifying their lengths to match the required electrical properties of the deleted wiring net. This conventional ATF repair method has worked well for traditional MCM-TF manufacturing. For tight ground rule MCM-TF products, however, a drawback of this conventional repair process is that product yield is adversely affected if the number of nets requiring repair exceeds the number of available repair nets on the TSM.
Referring again to FIG. 1, a typical pair of wiring nets 116, 118 are shown. For illustrative purposes, it is assumed that a short circuit exists between wiring nets 116, 118. The conventional repair process deletes the entire wiring nets 116, 118 by cutting wiring nets 116, 118 at C4 location 120. In this example, wiring nets 116, 118 are cut (also called deletes) at sites 122, 124, 126, 128, 130, 132, 134, 136, 138, 140, 142, 144, 146, and 148. The deleted wiring nets 116, 118 must be replaced using the TSM repair net (shown in FIG. 2A).
FIGS. 2A and 2B shows a typical TSM repair net 200 for the MCM of FIG. 1. In FIG. 2A, repair net 200 is made up of x-lines 202 and y-lines 204. As shown in FIG. 2B, within the gridwork of repair net 200 are C4 connections 206 for each chip 102, 104, 106, 108, 110, 112, and 114 mounted on MCM 100.
FIG. 2C shows an x-ray view of a five-layer MCM and FIG. 2D is a partial side view of MCM 100 illustrating the layered structure of MCM 100. In FIG. 2C, successive layers form MCM 100. Typical layers include ground layer 208, power layer 210, x-layer 212, and y-layer 214. An additional layer, top layer 216 (shown in FIG. 2D), contains repair net 200 and C4 connections 206. It is apparent from FIG. 2C that repair of an internal short circuit between any two x-layer lines or y-layer lines is a formidable task. For this reason, conventional repair processes deleted defective nets at the top layer 216.
As mentioned above, conventional ATF repair is based on full repair. That is, the entire internal structure of a defective net is removed at its C4 connections 206. An entirely new set of wiring is reconstructed using repair net 200 and connected to the C4 connections 206 on the TSM. These full repairs are necessary because frequently the location of the defect in the defective net is unclear and the construction of a new net is the only practical way to repair the defective net.
FIG. 2E illustrates a portion of a typical MCM before repair. In FIG. 2E, C4 connection 206 is connected to internal net 220 at via 238. X repair line 222 and Y repair lines 224, 226 are part of the top layer 216. Y repair lines 224, 226 are connected by Y repair line subway 236 using vias 228, 240.
The reconstruction of the net is normally accomplished by joining the segments of the repair lines with individual gold slugs bonded to the TSM of the repair through conventional lasersonic bonding processes. The gold slugs interconnect specific X and Y repair line segments to rebuild the net topography.
FIG. 2F illustrates the conventional repair process mentioned above. In FIG. 2F, when a short is found in internal net 220 it is completely disconnected from the circuit using external delete 230 between C4 connection 206 and via 238. This process is repeated at every other C4 connection location for internal net 220. To replace this deleted net, a portion of X repair line 222 and Y repair lines 224, 226 must be used. Conventionally, X repair line 222 and Y repair lines 224, 226 are cut using deletes 232. Then C4 connection 206 is connected to X repair line 222 and Y repair line 224 using gold slugs 234.
The drawback of this approach is that a relatively large number of repair lines are consumed for nets with multiple segments. As illustrated in FIG. 2F, an X repair line and a Y repair line were necessary to replace internal net 220. This results in fewer nets being repairable. An additional drawback of this conventional repair process is the scrapping of a part if an input/output (I/O) net is identified as defective. This is because conventional repair processes can only repair top-to-top signal nets while an I/O site is connected within the layers of the device.
Furthermore, because most defective nets run in the same general direction on the device, they require the use of the same top-surface repair lines. In such a case a part might be lost due to unroutability--insufficient repair lines to meet the repair requirements.